The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches data from external MySQL databases that may contain malicious instructions designed to influence the agent's behavior.\n- Ingestion points: Data results fetched via cursor.fetchmany() in scripts/query.py are returned directly to the agent's context.\n- Boundary markers: The output is presented as a plain text table without specific delimiters or instructions to the agent to disregard instructions found within the data.\n- Capability inventory: The agent typically possesses capabilities to interact with the file system and execute other tools based on the output it receives.\n- Sanitization: While the tool enforces structural safety (read-only mode, row limits), it does not sanitize the actual content of the retrieved data for prompt injection patterns.

mysql